Refined generic text detection: added partial matches for short phrases (e.g. 'company logo') while allowing branded logos

src/analyzer.py

@@ -52,11 +52,26 @@ class ImageAnalyzer:
                     is_poor = True
                     reason = "Filename pattern detected (slug)"
 
-                # B) Generic / Filler Text (Exact match check)
+                # B) Generic / Filler Text (Smart check)
+                # Flag if alt equals or contains generic terms.
+                # Exception: "Brand Name Logo" (usually > 2 words or specific)
+                # If word count is low (< 3) AND contains a generic term, it's likely poor.
+                # If it is EXACTLY a generic term, it is definitely poor.
                 generic_terms = {'image', 'photo', 'picture', 'logo', 'banner', 'icon', 'thumbnail', 'placeholder', 'img', 'spacer'}
-                if not is_poor and lower_alt in generic_terms:
-                     is_poor = True
-                     reason = "Generic filler text"
+                
+                if not is_poor:
+                    if lower_alt in generic_terms:
+                         # Exact match found (e.g., "logo", "image") -> Bad
+                         is_poor = True
+                         reason = "Generic filler text"
+                    else:
+                        # Check partial match for short phrases (likely "My Logo", "Chart Image")
+                        # If phrase is long (>= 3 words), it might be descriptive enough (e.g., "Eminent Tactiles Logo")
+                        # So we only flag partial generic matches if word count is small (< 3)
+                        has_generic = any(term in lower_alt.split() for term in generic_terms) 
+                        if has_generic and word_count < 3:
+                            is_poor = True
+                            reason = "Generic filler text (partial match)"
                 
                 # C) Too Short / Sparse (Lowest Priority)
                 if not is_poor: